Research On Part Defect Detection Method Based On Machine Vision

With the improvement of the degree of automation in modern industry,the traditional detection method has been unable to meet the needs of modern industry.So non-contact detection method arises at the historic moment and the detection method based on machine vision has become the key point of industrial detection research and application.The process of the detection based on machine vision is that the pictures tanken by the camera is transmitted to the computer by the image acquisition card,and then the image of the target parts is processed by the computer to get the geometric information of the parts.Based on this,this thesis main contents are as follows:Firstly,the hardware scheme of the system is determined according to the requirements of the project.Through the analysis of the camera,lens,light source and other aspects,the choice of each hardware is determined.In order to facilitate the detection of the typical defects to the target parts,this paper chooses the LED ring light source,which can meet the requirements of photographing the surface and contour of the target parts.Using the selection of each hardware in the system,a set of detection platform based on machine vision was successfully built,which can complete the process of the camera calibration and the shooting of the target parts.Secondly,the algorithm of image Mosaic is studied.Combined with the image of the target parts,the obtained image is preprocessed.In order to reduce the noise and retain the edge information of the target image,a method combining the improved median filter and bilateral filter is proposed,which lays a good foundation for image Mosaic.Then the SIFT algorithm is selected to match the image,and the accurate image matching transformation matrix is obtained.Finally,the weighted average method is used for image fusion.The results show that the image Mosaic algorithm is effective.Third,Different detection methods based on two different defects are proposed.The defects on the target parts are divided into burr type protruding defect and pit type and scratch type surface defect,and the two defects are detected by different methods.The burr type protruding defect mainly adopts the method of comparing the binarization image of the standard part with the binarization image of the target part.Surface defects such as pits and scratches are extracted mainly by Grab Cut image segmentation.At the same time,an improved Laplace operator image enhancement method and an improved iterative method threshold segmentation algorithm are proposed to achieve effective segmentation between defects and background.Finally,the image edge detection algorithm and defect size detection method are studied.In order to reduce the measurement error,a method is used to fit the sub-pixel image which is combined the improved Canny operator and the improved least square method.An irregular figure size detection method is also proposed,which can better detect the area size and maximum length of the defect to meet the requirements of the inspection of the defect size.Finally,the visual inspection system of metal parts is designed,which can meet the requirements of the detection of metal parts and the measurement of their defect size.